The Influence of Deliberate Practice on Musical Achievement: A Meta-Analysis

k=13 studies, N=788 participants (screened from a preliminary corpus of 102 studies)

Musicians, mostly instrumentalists, spanning student to advanced/professional levels; studies published 1992-2012 including grey literature and dissertations.

k=13, N=788. Deliberately NARROW inclusion: instrument-specific task-relevant practice paired with objective/validated achievement measures. Hunter-Schmidt psychometric meta-analysis correcting each study for sampling error and measurement unreliability. Homogeneous: I2=0.00%, Q(12)=8.19, p=.77; 81.2% of the variance in corrected correlations is attributable to artifacts. Funnel plot near-symmetrical around rc=.61 with one outlier, so no trim-and-fill or PET-PEESE correction was applied or needed by the authors' reading. Cross-sectional and correlational throughout — no experimental or longitudinal data. Independence caveat: three of the thirteen included effect sizes are the meta-analysts' own studies (Lehmann & Ericsson 1996, Kopiez & Lee 2008, Kopiez et al. 2012, ~10.8% of total weight), and co-author A. C. Lehmann is a long-standing Ericsson collaborator; the paper is framed throughout as a defence of the deliberate-practice position against Hambrick/Macnamara. The declared conflict-of-interest statement is "none."

The strongest pro-practice meta-analysis, and therefore the best-case number: even with a narrow DP-favourable definition, objective outcomes, and full artifact correction, practice tops out at rc=.61, leaving roughly 60%+ of achievement variance outside practice hours. Quote the uncorrected r=.54 for comparability with Macnamara. Note that the authors themselves reject the variance-explained reading in both directions — "it is incorrect to interpret our findings (rc = 0.61) as evidence that DP explains 36% of the variance in attained music performance" — so the 60%-unexplained framing is our inference from their coefficient, not their claim. Music is the domain where practice matters most because it is highly predictable and structured; this ceiling should not be generalized to less-structured domains (Macnamara et al. 2014 report education 4%, professions <1%).

None controlled. Undercut by Mosing et al. (2014): within MZ twin pairs discordant for practice, no music-ability difference. Practice is ~40-70% heritable, so much of this correlation is gene-environment correlation, not causation.

Numerically consistent with Hambrick et al. (2014), whose comparable corrected estimate on 8 studies was rc=.52 [.43, .64] — overlapping confidence intervals — and with Macnamara et al. (2014) music (21-23% of variance) once the uncorrected r is used. Cited by the Ericsson camp as rebuttal; the gap is correction method plus inclusion strictness, not disagreement about raw data.